Lots of $$ and lots of potential...
The problem is that 100 trades isn't enough to come up with any significant conclusions about a system. Some brokers allow you to download up to 7 years of data. Try backtesting it on a few more years worth of data.
There are people here using systems tested on the same data I used, which is Alpari. The formula for statistical signifigance is 1/root n, n being the number of cases. On a longer term method the n must be over 400 cases to get +/- 5% statistical accuracy, plus the data sets available to us suck. We're in retail. We're poor bastards or we wouldn't be here posting on forums and looking for methods. I'm developingmethods! Going from 1000USD to 8000USD in 2 years was a f#$cking accomplishment! The data I need to test this accurately costs 1800EUR from Olsen. I can't just come up with 1800EUR! That's why I put this out there. I want help testing it!
I can trade okay but I have limited capital! I work like a dog for a living and I count my trading money as money spent...
I didn't say that this actually will work over time, I actually said I don't know. I was asking for help with optimizing and testing it...
It's an idea.
We shift by whatever number is recently optimal
So you are curve fitting, which is ok....but how do you determine what is "recently optimal"? The past 3 months, the past 3 years?
I took at cursory look at this, using a 90ema and a 30ema shifted 20, on a 4 hour chart, and it actually looks pretty good.
So, how exactly do you want us to help you?
I would really like longer term testing of the basic idea, but I'm just okay with a computer. Really, if anyone has a few different moving average EAs I'd like to test it with different EAs on different sets of data. I noticed in Simba's thread that they found some larger data sets, so I'm going to get those, but if anyone has any other reasonably good MA cross EAs, I'd appreciate those as well. That way I'm not just counting on the reliability of a single EA, this would confirm a lot more...
I don't think this is going to work, it's way too easy to overoptimize...
I've been going in circles trying to find a way to do it that tests well in and out of sample, and anything that works for a few years, doesn't work for the next few. It will break about even out of sample and perform amazingly on the data it's tested on, no matter what I do with it. I may need to filter the trades more, etc., but it's becoming quite time-consuming...
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
Here's a method you guys might be interested in. It has made me money in real time, but I need to test it to make sure it's not some fleeting B.S.
The work is based on Tom Demark's "The New Science of Technical Analysis", however it is not applied in the manner he presents it. This is an entirely new idea. Shifted MAs have been around a long time, but here's my take on it.
When we observe past price action, we generally tend to assume the obvious, that is, if the fast MA is above the slow MA the trend is long. This is not necessarily correct, and in many cases, is counterintuitive. The point is that we can play a zero-sum game with moving average cross system users that exploits the weaknesses of their systems.
Fast/slow moving average systems tend to lose money. They are generally on the wrong side of the market, at least in the short-term. The reason is that slow traders like import/export companies and multi-national corporations use these simple methods for hedging in the long-term. The swing trader will generally be on the wrong side when applying these methods to shorter-term analysis.
Instead of Fibonacci, I have found that GBP/USD tends to follow a one third, two thirds cycle. In other words, It will lose 1/3 as it advances 2/3, in a long term uptrend. I chose a 90 EMA, or base period, unshifted, and a 30 EMA shifted 20 periods, or 2/3, for these tests.
I'm talking about methods that succeed over years on the long term charts that fail on a daily or 4 hour chart. I'm making them work on shorter charts. Be forewarned, these settings probably won't work in the future, I'm working on ways of finding optimal settings dynamically. Read Demark, Chapter 6, and he sets up this idea. Here's my take:
1. We use a longer period exponential moving average, somewhere between 80 and 90 on a 4 hour chart, and a shorter period, around 15-25. We only shift the second MA, but shift it at a value almost its own value, within a range.
2. I am not fixated on fibonacci numbers, if I was I'd be broke by now. We shift by whatever number is recently optimal.
3. We buywhen the slow MA crosses abovethe fast MA.
4. We sell when the slow MA crosses below the fast MA.
5. Read #3 and #4 again.
6. Look at the results on Cable backtesting...
Please help to test and optimize this, it's really promising...